Journal of Seasonal ice-speed variations in 10 marine- terminating outlet along the coast of Prudhoe Land, northwestern Greenland

Paper Daiki Sakakibara1,2 and Shin Sugiyama2,1

Cite this article: Sakakibara D, Sugiyama S 1Arctic Research Center, Hokkaido University, Kita-21, Nishi-11, Kita-ku, Sapporo, 001-0021, Japan and 2Institute of (2020). Seasonal ice-speed variations in 10 Low Temperature Science, Hokkaido University, Kita-19, Nishi-8, Kita-ku, Sapporo, 060-0819, Japan marine-terminating outlet glaciers along the coast of Prudhoe Land, northwestern Greenland. Journal of Glaciology 66(255), Abstract 25–34. https://doi.org/10.1017/jog.2019.81 We present a 3-year record of seasonal variations in ice speed and frontal of 10 marine- Received: 18 April 2019 terminating outlet glaciers along the coast of Prudhoe Land in northwestern Greenland. The Revised: 14 October 2019 glaciers showed seasonal speedup initiated between late May and early June, and terminated Accepted: 14 October 2019 between late June and early July. Ice speed subsequently decreased from July to September. First published online: 13 November 2019 The timing of the speedup coincided with the onset of the air temperature rise to above freezing, Key words: suggesting an influence of meltwater availability on the dynamics. No clear relationship Arctic glaciology; glacier monitoring; ice was found between the speedup and the terminus position or the sea-ice/ice-mélange conditions. velocity; remote sensing These results suggest that the meltwater input to the glacier bed triggered the summer speedup. The excess of summer speed (June–August) over the mean for the rest of the year accounted for Author for correspondence: – Daiki Sakakibara, E-mail: sakakibara@pop. 0.5 13% of the annual ice motion. Several glaciers showed seasonal frontal variations, i.e. retreat lowtem.hokudai.ac.jp in summer and advance in winter. This was not due to ice-speed variations, but was driven by seasonal variations in frontal ablation. The results demonstrate the dominant effect of glacier surface melting on the seasonal speedup, and the importance of seasonal speed patterns on longer-term ice motion of marine-terminating outlet glaciers in Greenland.

1. Introduction Mass loss from the Greenland has contributed to global sea level rise at a mean rate of − 0.47 mm a 1 during the period from 1991 to 2015 (van den Broeke and others, 2016). Since ∼40% of the mass loss was due to changes in ice discharge from marine-terminating outlet glaciers (van den Broeke and others, 2016), a future projection of ice discharge from the Greenland ice sheet is crucial for understanding the magnitude and timing of global sea level rise. Ice discharge from the ice sheet is dependent on the ice speed of marine-terminating outlet glaciers. Glaciers in Greenland had accelerated substantially in the early 2000s, and ice speed has been kept at the accelerated level since 2005 (e.g. Howat and others, 2008; Moon and others, 2012; King and others, 2018; Mankoff and others, 2019). Observed speed changes are inhomogeneous in space and time; thus further investigations and insights are needed to pro- ject future changes in marine-terminating outlet glaciers in view of our changing climate. Recent progress in satellite remote sensing techniques contributes to the studies of seasonal speed patterns of marine-terminating glaciers in Greenland (Howat and others, 2010; Moon and others, 2014, 2015; Vijay and others, 2019). Reported data indicate significant summer acceleration, suggesting the importance of high-frequency ice-speed measurements to the quantification of annual ice discharge. Sole and others (2011) reported the ∼1% contribution of summer speedup to annual ice motion of a marine-terminating outlet glacier in southwest Greenland. However, as the study site was far inland from the at >1200 m a.s.l., the importance of summer speedup to glacier dynamics in the vicinity of the terminus remains poorly understood. Seasonal speed variations are essential also for understanding the response of ice dynamics to external forcings (Howat and others, 2010; Carr and others, 2013; Moon and others, 2014; Lemos and others, 2018; Vijay and others, 2019). Moon and others (2014) reported the sea- sonal speed patterns of 55 marine-terminating glaciers along the coast of northwest, west, southeast and southwest Greenland. This study classifies seasonal speed patterns into three types, depending on their sensitivity to the terminus position, meltwater production and the subglacial hydrological system. Seasonal speed patterns in northwestern Greenland were correlated with meltwater production, suggesting a higher sensitivity of the glacier dynamics in this region to climate variability (Moon and others, 2014). More recently, higher frequency © The Author(s) 2019. This is an Open Access ice-speed measurements were performed on 45 glaciers over broader areas in Greenland (Vijay article, distributed under the terms of the Creative Commons Attribution licence (http:// and others, 2019). This study shows the concurrence of summer speedup and the onset of gla- creativecommons.org/licenses/by/4.0/), which cier surface melting, which essentially agrees with the conclusion by Moon and others (2014). permits unrestricted re-use, distribution, and These studies indicate the importance of expanding the spatial and temporal coverage of these reproduction in any medium, provided the data to include the investigation of future responses of Greenlandic outlet glaciers to changing original work is properly cited. climate. Seasonal speed variations in marine-terminating glaciers are influenced by several factors such as subglacial hydraulic conditions, surface melting and drainage into the glacier bed, ter- cambridge.org/jog minus variations and sea-ice/ice-mélange conditions (e.g. Howat and others, 2010; Carr and

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use. 26 Daiki Sakakibara and Shin Sugiyama

others, 2013; Moon and others, 2014; Pimentel and others, 2017). ∼20 km wide Inglefield Bredning. Hubbard, Bowdoin, Numerical modeling and observational studies indicate the Verhoeff, Morris Jesup and Diebitsch Glaciers are located in the importance of the glacier terminus position in ice dynamics western half of the study area, where each individual fjord con- (Howat and others, 2008; Joughin and others, 2008a, 2008b; nects to western part of Inglefield Bredning or directly to northern Nick and others, 2009). Observations in southeast Greenland Baffin Bay. The ice speed near the termini of these glaciers ranges − have shown that resistive stress acting on the terminus is dimin- from 90 to 400 m a 1 (Sakakibara and Sugiyama, 2018), which is ished when a marine-terminating glacier retreats, a process smaller than the speed of the glaciers in the eastern half. Heilprin, which is significant enough to cause flow acceleration (Howat Farquhar, Bowdoin and Diebitsch Glaciers showed a substantial and others, 2008). Numerical experiments on Helheim Glacier speedup and retreated rapidly after 2000 (Sakakibara and confirmed the critical effect on glacier dynamics of bed geometry Sugiyama, 2018). near the terminus (Nick and others, 2009). Sea ice and ice mélange in front of a marine-terminating glacier are thought to have a buttressing effect on the glacier front (Joughin and others, 3. Data and methods 2008a; Amundson and others, 2010). Therefore, the melting of To measure the glacier terminus position, ice speed and sea-ice/ sea ice and ice mélange has the potential to reduce resistive stress ice-mélange conditions from 2014 to 2016, we used satellite acting on the glacier and cause summer speedup. The meltwater images of the Landsat 8 Operational Land Imager (OLI) (level input to the glacier bed and the evolution of a subglacial drainage 1T, with 15/30 m resolutions) distributed by the United States system also influence seasonal speed variations (Joughin and Geological Survey. Because the OLI is an optical sensor, the mea- others, 2008b; Schoof, 2010; Sole and others, 2011). In early sum- surements were performed when sunlight was available between mer, surface melt increases and water drains to the bed, where a March and October. subglacial drainage system is not sufficiently developed. In such a We obtained hourly air temperatures at Mittarfik Qaanaaq situation, subglacial water pressure increases and is (Qaanaaq Airport; 77.48°N, 69.38°W; 16 m a.s.l.) located within enhanced. The drainage system develops further as a greater 30–84 km of the glacier termini, from the National Oceanic and amount of meltwater drains, resulting in a drop in subglacial Atmospheric Administration. The temperature dataset was uti- water pressure and a decrease in ice speed (e.g. Bartholomew lized to calculate mean air temperature during a period for each and others, 2010; Sundal and others, 2011). These mechanisms ice-speed measurement. The Qaanaaq temperature showed a have been proposed and tested at some of the glaciers in good correlation (r2 = 0.94, p < 0.001) with that recorded at Greenland, but processes affecting seasonal speed variations are Thule Airbase located 108 km south from Qaanaaq. Thus, we not yet fully understood. assume the data well represent air temperature over the study area. To investigate seasonal speed variations of marine-terminating outlet glaciers, we processed satellite images for high-frequency ice-speed variations in 10 marine-terminating outlet glaciers 3.1. Ice speed along the coast of the Prudhoe Land, northwestern Greenland. We compared observed ice-speed changes in years 2014, 2015 The ice speed was measured by automatically tracking glacier sur- and 2016 with the terminus position, sea-ice/ice-mélange condi- face features on temporally separated satellite-image pairs tions and air temperature in order to discover the source of sea- (Scambos and others, 1992; Sakakibara and Sugiyama, 2014, sonal speed variations. The contribution of summer speedup to 2018). We used Landsat 8 OLI images obtained from 2014 to annual ice motion was quantified to evaluate its importance to 2016 for this purpose. These images were acquired with temporal ice discharge from the glaciers. We also computed seasonal varia- separations of 16–48 days and study glaciers were covered by mul- tions in frontal ablation, by comparing the ice speed and terminus tiple tracks, thus enabling high-frequency speed measurements to positions obtained in this study. be made. The frequency of the ice-speed measurement at each glacier was 44–196 times per year, depending on the availability of the scene. Glacier surface displacement was measured with 2. Study site an automatic image matching scheme known as the orientation We focused on 10 marine-terminating outlet glaciers located correlation method in the frequency domain (CCF-O) (Heid along the coast of Prudhoe Land in northwestern Greenland and Kääb, 2012). The correlation window size was 30 pixels, (77.53–77.94°N and 66.04–71.63°W) (Fig. 1). This area has and the Landsat images were co-registered using displacement been a central part of previous studies on the dynamics of outlet vectors obtained in off-ice areas. Measurements were taken over glaciers (Porter and others, 2014; Sugiyama and others, 2015; areas near the glacier termini, and we sampled speed at the glacier Tsutaki and others, 2016; Sakakibara and Sugiyama, 2018), but center at a distance of 0.7–2.4 km from the termini in 2014. The not covered by past studies of seasonal speed variations (Moon distance from the terminus to the sampling site for each glacier and others, 2014, 2015; Vijay and others, 2019). A number of out- was approximately equal to the half of the glacier front width let glaciers, having a front width between 1.5 km (Sharp Glacier) (Moon and others, 2014). The magnitude of summer speedup and 5.0 km (Tracy Glacier), are located within a relatively small was dependent on the distance from the glacier front, but the tim- area. Glaciers in the study area have been retreating since the ing of the peak speed and seasonal patterns was not influenced by early 2000s and some of the glaciers have shown significant accel- the sampling location. Displacement vectors obtained with a low eration during this period (Sakakibara and Sugiyama, 2018). signal-to-noise ratio (<30) were excluded, and remaining vectors Heilprin and Tracy Glaciers are the two largest glaciers in this were filtered using directional and median low-pass filters region, having a drainage basin of 1.04 × 104 km2 in total (Rignot (Sakakibara and Sugiyama, 2014). Ice-speed data at the sampled and Kanagaratnam, 2006). Tracy Glacier retreated at a rate of locations were smoothed by taking the median values of 7 by 7 − − 200 m a 1 and accelerated from 810 to 1740 m a 1 during the neighboring pixels (Howat and others, 2010). Ambiguities due period from 1980s to 2014 (Sakakibara and Sugiyama, 2018), to the uncertainty in the cross-correlation peak and − which was the greatest change recorded among the study glaciers. co-registration errors ranged from 6 to 300 m a 1 with a mean − Farquhar, Melville and Sharp Glaciers flow at a rate more than value of 43 m a 1, depending on image quality and length of tem- − 400 m a 1 (Sakakibara and Sugiyama, 2018). These five glaciers poral separation. After the filtering and smoothing procedures, are located in the eastern half of the study area, draining into a measurement error was mostly associated with the co-registration

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use. Journal of Glaciology 27

Fig. 1. The locations of the study glaciers along the coast of Prudhoe Land. The background is a true color mosaic Landsat image acquired on 9 July 2014. The inset shows the location of the study area in Greenland.

of image pairs. Thus, the uncertainty was not influenced by sea- glacier center. The conversion factor was derived for each glacier sonal changes in glacier surface conditions. by computing a mean speed across the glacier using all available measurements. We assume ice speed was uniformly distributed from the surface to the bottom (perfect sliding). Uncertainty in 3.2. Terminus position and sea-ice/ice-mélange conditions the calculated frontal ablation was estimated as 15 to 24 m from Glacier terminus positions were mapped from 2014 to 2016, using the error ranges assumed in the terminus position and ice-speed false color composite images generated by assigning the colors measurements. blue, green and red to bands 4, 5 and 6 of the Landsat 8 images, respectively (McNabb and Hock, 2014). Relatively coarse false color composite images were used for this purpose because the 4. Results boundary between a glacier terminus and sea ice was more clearly observed than panchromatic images available in a higher reso- 4.1. Ice speed lution. In each image, the glacier frontal margins were delineated The study glaciers showed clear seasonal speed variations (Fig. 3). manually using the QGIS geographic information system soft- For example, the ice speed at Tracy Glacier decreased slightly − ware. Mean displacement of the terminus position was computed from 2200 to 2100 m a 1 during the period from March to May by dividing the areal change near the frontal margin by the width 2015 (Fig. 3e). The glacier progressively accelerated to − of the terminus (Moon and Joughin, 2008). Relative positioning 2500 m a 1 from late May to early July, followed by gradual decel- − errors introduced during the process of overlapping the Landsat eration to 2200 m a 1 from July to September. A similar seasonal 8 images was ±8 m (Sakakibara and Sugiyama, 2018). This speed pattern was observed at Bowdoin Glacier. The ice speed at − value was derived as the co-registration error in the ice-speed Bowdoin Glacier was within the range of 300 to 400 m a 1 during measurement, by computing the root-mean-square of displace- the period from March to May 2015 (Fig. 4e). Ice speed increased − ment obtained in off-ice areas. from 400 to 600 m a 1 from late May to early July, and decreased − Sea ice and ice mélange conditions near the termini were to 350 m a 1 from July to September. Early summer speedup and determined using a method proposed by Moon and others late summer slowdown were commonly observed at the other (2015). In every image, sea-ice/ice-mélange concentration and ∼ study glaciers (Fig. 4). In general, ice speed showed only minor potential rigidity within 5 km from each glacier terminus were changes from March to May. Acceleration typically began classified as likely to be rigid (rigid), potentially rigid (mixed) between late May and early June, followed by the achievement or unlikely rigid (open) (Moon and others, 2015) (example is ‘ ’ of a peak speed between late June and early July. Ice speed subse- shown in Fig. 2). The condition was classified as rigid when quently slowed down from July to September. the region in front of the glacier was completely or near- ‘ ’ The timing of the seasonal speedups and slowdowns is com- completely covered by sea ice/ice mélange. An open classification parable in the study glaciers regardless of size and location. was given for open water. When the region was covered with a However, the magnitude of summer speedup was significantly substantial amount of ice but also with open water, the condition ‘ ’ different for each glacier. The most significant speedup was was classified as mixed (Moon and others, 2015). observed at Farquhar Glacier, where ice speed increased by − more than 300 m a 1. The magnitude of the speedup at other gla- − ciers ranged between 54 and 278 m a 1 corresponding to a range 3.3. Frontal ablation between 8 and 226% increase relative to the speed before the Frontal ablation (sum of calving and subaqueous melting) was acceleration. Seasonal speed patterns do not always follow the calculated by subtracting the terminus displacement from ice general trend described above. For example, ice speed at motion over a period from March 2014 to September 2016. Heilprin Glacier showed no significant change from May to July Mean ice speed across the glacier was estimated as 55–90% in 2014 (Fig. 3a). Speedup was unclear also at Farquhar Glacier (mean: 73%) of the speed obtained at the sampling site at the in 2014 (Fig. 3g).

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use. 28 Daiki Sakakibara and Shin Sugiyama

Fig. 2. False color mosaic Landsat images used for categorization of fjord conditions. The examples show conditions in front of Tracy Glacier, which were cate- gorized as (a) rigid, (b) mixed and (c) open sea ice.

4.2. Terminus position and sea-ice/ice-mélange conditions To investigate a link between ice speed and atmospheric con- ditions, the ice-speed time series data were plotted against air Some of the glaciers showed substantial sudden retreat due to temperature averaged over the same period (Fig. 5). This analysis large calving events, typically by more than 100 m within the was performed for the data in 2015 because ice speed was mea- interval of satellite images ranging between 2 and 23 days. For sured most frequently for that year out of the study years. The example, Heilprin Glacier retreated by 200 m in June 2014 plots demonstrate an influence of air temperature on the speed (Fig. 3a) and in April 2015, and 110 m in May 2015 (Fig. 3b). of the studied glaciers. Similarly, sporadic major calving events were observed at Tracy For example, ice speed at Heilprin Glacier was fairly stable (Figs 3d to f) and Melville Glaciers (Figs 3j to l). Farquhar − within 1300–1600 m a 1 under freezing temperature conditions (Figs 3g to i) and Bowdoin Glaciers (Figs 3d to f) experienced (arrow 1 in Fig. 5a). This corresponds to a period from March slightly smaller events (>50 m retreat). These sudden retreats to May (Figs 3b, 3q). About one month after hourly mean tem- were followed by gradual advances lasting for a few months. perature exceeded 0°C (vertical line in Fig. 3q), ice speed began Some other glaciers showed seasonal variations in the terminus to increase, and further accelerated in the course of the tempera- position. For example, Morris Jesup Glacier advanced by 30 m ture rise. Ice speed and temperature showed a positive correlation from May to June and retreated by 130 m from early July to until speed reached the maximum in early July (arrow 2 in September in 2014 (Fig. 4j). Verhoeff and Diebitsch Glaciers Fig. 5a). Similar speed-temperature relationships were observed showed similar frontal variations (Figs 4g to i and 4m to o). at Tracy, Farquhar, Bowdoin and Verhoeff Glaciers (Figs 5b to The change in the trend from advance to retreat occurred typic- c and 5g to h). The other glaciers showed some unique character- ally in June or July. istics, but timing and trend of the speedup were similar to the Sea ice and ice mélange disappeared at most of the glacier relationship discussed above, i.e. insignificant speed change fronts in late June to early July (Figs 3, 4). The timing of the when temperatures are below zero, and acceleration to the sum- sea-ice/ice-mélange disappearance was consistent over the three mer maximum speed as temperatures increase. The relationship years studied. opened earlier at Morris Jesup and implies that the summer speedup was triggered by meltwater pro- Diebitsch Glaciers, both situated in the western part of the duction and enhanced as the melt rate increased. Based on these study area (Figs 4j to o). Because the fjords of these two glaciers observations, we assume that subglacial water pressure increased are directly connected to Baffin Bay, sea ice and ice mélange were due to meltwater penetration into the glacier bed, and the glacier released earlier in summer that is from mid-April to early June. accelerated due to enhanced basal sliding (e.g. Iken, 1972; By contrast, sea ice and ice mélange stayed longer after the Andrews and others, 2014; Ryser and others, 2014; Vijay and onset of melting in Inglefield Bredning. At the fronts of others, 2019). Heilprin, Tracy, Farquhar and Melville Glaciers, mixed conditions After peak speed was attained at Heilprin Glacier, ice motion lasted for more than two weeks before the glacier front became rapidly decelerated while air temperature remained high (arrow 3 ice-free. No clear relationship was found between the sea-ice/ in Fig. 5a). Ice speed remained low until the end of the summer, ice-mélange conditions and the large calving events observed at resulting in a hysteresis loop between ice speed and temperature. these glaciers. A similar drop in ice speed from July to August was observed at all glaciers (Fig. 5). The glaciers decelerated in the middle of sum- 5. Discussion mer when meltwater was abundant, which we attribute to the sea- sonal evolution of the subglacial drainage system. Observations at 5.1. Mechanism of seasonal speed variations alpine glaciers have demonstrated that an inefficiently linked cav- Our data reveal details of seasonal speed variations along the coast ity system evolves into an efficient conduit drainage system over of Prudhoe Land. All the study glaciers show speedup in summer. the course of a melt season (e.g. Iken and Bindschadler, 1986; The speedup initiated between late May and early June, speed Fountain and Walder, 1998). After the efficient drainage system gradually increased until early July and then it slowed down in is established, subglacial water pressure drops and glacier speed the period from July to September. The magnitude of the seasonal becomes unaffected by the amount of meltwater input to the − speedup was a few 100 m a 1 which corresponds to 8–226% bed (e.g. Sugiyama and Gudmundsson, 2003; Schoof, 2010). increase relative to pre-speedup conditions (Figs 3, 4). The This interpretation is consistent with the speedup mechanism observed seasonal speed pattern and magnitude of the speed described previously. change are similar to those reported in the Melville Bay region At Melville Glacier, ice speed abruptly increased to − in northwestern Greenland (74.92–76.08°N) (Moon and others, >1300 m a 1 when the air temperature was still below 0°C 2014). Thus, the observed seasonal speed pattern appears to be (Fig. 5d). This speedup occurred in May, when the fjord was a ubiquitous feature across northwestern Greenland, and the ice- still covered with sea ice, immediately after the sudden terminus speed variations are likely driven by a common mechanism. retreat of ∼300 m due to a large calving event (Fig. 3k). Except

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use. Journal of Glaciology 29

Fig. 3. (a–o) Ice speed (red), terminus position (blue) and sea-ice/ice-mélange conditions (magenta = rigid, gray = mixed and cyan = open) for five study glaciersin the eastern part of the study area. Red lines and associated shaded bands indicate daily mean speed and the std dev., respectively. The distance shown next to the glacier names indicates the location of ice-speed sampling as measured from the 2014 termini. (p–r) Daily (orange) and hourly (yellow) air temperatures, and the sum of the positive degree hours (PDH) (blue) at Qaanaaq Airport. The vertical orange lines show the first day of positive hourly temperature in each year.

for this incident, no clear relationship was observed between speedup. Sea ice and ice mélange disappeared typically in early frontal variations and glacier dynamics. A sudden retreat of July when summer speedup ceased (Figs 3, 4). >200 m was observed frequently at Heilprin (Figs 3a, b), Tracy Based on the discussion above, we conclude that summer (Figs 3d to f) and other glaciers, but ice speed was unaffected speedup observed at the glaciers in Prudhoe Land was driven by these events. Seasonal variations in the front position were by subglacial water pressure which increased due to the meltwater also not relevant to the change in speed. Morris Jesup, Verhoeff input to the bed. The glaciers decelerated in midsummer when an and Diebitsch Glaciers began retreat in June or July, whereas sum- efficient drainage system developed and water pressure dropped. mer speedup of these glaciers began already in May. Therefore, This result is consistent with a recent study on other glaciers in reduction in resistive stress due to terminus retreat (Joughin Greenland reporting the primary influence of surface meltwater and others, 2008a, 2014) is not the source of the observed sea- production and associated subglacial hydraulic conditions on a sonal speed variations. We also found no link between sea-ice/ seasonal ice-speed change (Vijay and others, 2019). The effect ice-mélange disappearance from the fjords and the onset of the of the glacier front position on the ice speed is limited, although

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use. 30 Daiki Sakakibara and Shin Sugiyama

Fig. 4. (a–o) Ice speed (red), terminus position (blue) and sea-ice/ice-mélange conditions (magenta = rigid, gray = mixed and cyan = open) for five study glaciersin the western part of the study area. Red lines and associated shaded bands indicate daily mean speed and the std dev., respectively. The distance shown next to the glacier names indicates the location of ice-speed sampling as measured from the 2014 termini. (p–r) Daily (orange) and hourly (yellow) air temperatures, and the sum of the PDH (blue) at Qaanaaq Airport. The vertical orange lines show the first day of positive hourly temperature in each year.

sudden retreat due to a large calving event sometimes triggered difference between summer speed (June–August) and ‘base glacier acceleration. speed’ defined by the mean over the rest of the year (September–May of the following year). In other words, summer speedup refers to the additional displacement that occurs on top 5.2 Contribution of summer speedup to glacier motion of the base flow. The magnitude of the summer speedups was As demonstrated in recent studies (Moon and others, 2014, compared with annual ice motion, using data from September 2015), seasonal speed patterns are crucial for quantifying the 2014 to August 2016 (Fig. 6). annual discharge of marine-terminating outlet glaciers in The contribution of the summer speedup to annual motion of Greenland. To investigate the relative importance of the fast-flowing Heilprin, Tracy, Farquhar and Melville Glaciers, − observed summer speedup in longer-term ice discharge, we which flow at a rate >1000 m a 1, was within the range of − quantified the contribution of summer speedup to annual ice 24–62 m a 1 (2–5%) (Fig. 6a). Summer speedup contribution − motion. We define the magnitude of summer speedup as the of the other six glaciers ranged from 2 to 23 m a 1 (0.5–13%)

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use. Journal of Glaciology 31

abc d

ef gh

ij

Fig. 5. Scatter plots of ice speed and air temperature at Qaanaaq Airport averaged over each ice-speed measurement period. The data covers a period from April to October 2015.

Fig. 6. The contribution of summer speedup (June–August) to mean ice speed from September 2014 to August 2016 for (a) Heilprin, Tracy, Farquhar and Melville Glaciers and (b) Sharp, Hubbard, Bowdoin, Verhoeff, Morris Jesup and Diebitsch Glaciers. The numbers indicate the relative contributions of the summer speedup in %. (c) Scatter plots of the summer speedup and its relative contribution vs the mean ice speed.

(Fig. 6b). Summer increase in total ice discharge from the accounted for more than 10% of annual ice motion. The numbers Greenland ice sheet is reported as ∼6% of the annual mean obtained for the slow-flowing glacier are similar to those reported (King and others, 2018), which implies a 2–3% contribution of for a land-terminating glacier in western Greenland (6–14%) summer speedup to annual ice discharge. This estimate is consist- (Bartholomew and others, 2010). In the previous studies, the ent with our results obtained for fast-flowing glaciers. Our results magnitude of summer speedup of marine-terminating glaciers demonstrate that the impact of summer speedup on annual ice was greater near the terminus than inland (Sole and others, motion was greater at relatively slow-flowing glaciers, whereas lar- 2011; Joughin and others, 2014). Therefore, it is important to ger magnitude of speedup was observed at fast-flowing glaciers understand seasonal speed patterns near the glacier termini to (Fig. 6c). Seasonal speed variation was particularly important at accurately quantify annual ice discharge from marine-terminating Verhoeff and Morris Jesup Glaciers, where summer speedup glaciers.

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use. 32 Daiki Sakakibara and Shin Sugiyama

position, i.e. retreat in summer and advance during the rest of the year (Fig. 7b). Forward ice motion was greater in summer, but enhanced frontal ablation was the dominant factor affecting change in the terminus position. Similar trends in the frontal ablation were observed in Melville, Verhoeff, Morris Jesup and Diebitsch Glaciers (Figs s2c, f, g and h). The contribution of the summer frontal ablation (June– August) to the total ablation varied for each glacier (Fig. 8). At Heilprin Glacier, summer ablation accounts for 22% − (140 m a 1) of the ablation that occurred from September 2014 to August 2016, indicating a near-uniform distribution over the period of a year. The contribution was slightly greater at Tracy, Farquhar, Melville, Sharp and Hubbard Glaciers, ranging from − 33 to 41% (60–570 m a 1). Around the half of the annual frontal ablation occurred during summer at Bowdoin, Verhoeff, Morris − Jesup and Diebitsch Glaciers (80–180 m a 1)(Figs 8a, b). Summer frontal ablation is generally correlated with annual abla- Fig. 7. Cumulative ice motion near the glacier front (blue), ice front displacement tion, but its relative contribution decreases as annual ablation (positive number shows retreat) (red) and cumulative frontal ablation obtained as increases (Fig. 8c). the sum of the ice motion and the frontal displacement (green) from 2014 to 2016 The first group, glaciers with year-round distributed frontal for (a) Tracy and (b) Bowdoin Glacier. ablation, was characterized by fast ice flow and large-glacier size. We assume that the influence of summer speedup and sea-ice/ice-mélange melting on calving was less significant at 5.3. Seasonal variations in the frontal ablation this type of glacier. This is because the relative contribution of The high-frequency ice speed and terminus position data enabled summer speedup to annual ice motion is small at these glaciers us to quantify temporal variations in frontal ablation. Figure 7 (Fig. 6). It is also likely that the buttressing effect of sea ice and shows cumulative forward ice motion near the glacier front, ter- ice mélange was insufficient to stabilize the fast-flowing glacier minus position (positive number shows retreat) and cumulative front. The greater contribution of summer frontal ablation at frontal ablation of Tracy and Bowdoin Glaciers. The frontal abla- other glaciers can be explained by calving enhanced by speedup tion of Tracy Glacier occurred at a nearly constant rate through- and the loss of sea ice/ice mélange, which play a more important out the study period, except for several calving events represented role in the calving of smaller and slow-flowing glaciers. by the sudden retreat in May and July 2015, and May 2016 We also suspect an influence of submarine melting on the sea- (Fig. 7a). These large calving events occurred irrespective of sonal variations in the frontal ablation. Fjord depth at the glacier sea-ice/ice-mélange conditions (Figs 3e, f). Frontal ablation at front is important for submarine melting because it affects the Heilprin and Farquhar Glaciers showed seasonal patterns similar intrusion of warm Atlantic water (AW) (Porter and others, to that at Tracy Glacier (Figs s2a, b). Large calving events are 2014; Rignot and others, 2016). According to the glacier bed ele- more frequent during summer periods, but the glacier front stead- vation measured in 2014 (CReSIS, 2016), the fjords are deeper ily loses ice by frontal ablation throughout the year. In contrast to than 300 m below sea level at the front of Heilprin, Tracy and these three glaciers, frontal ablation at Bowdoin Glacier was Farquhar Glaciers. AW intrudes into such deep fjords and pro- greater in summer than during the other period (Fig. 7b). vides heat for submarine melting throughout the year (Porter Ablation during a period from June to August (25% period of a and others, 2014). In contrast to these large glaciers, Bowdoin, year) accounted for 59% of the total ablation which occurred Verhoeff, Morris Jesup and Diebitsch Glaciers flow into fjords from September 2014 to August 2016. This seasonality in the shallower than 200 m (Sugiyama and others, 2015; OMG mission, frontal ablation caused seasonal variations in the terminus 2016). A submarine melt rate increases in summer because of

Fig. 8. Frontal ablation rate for (a) Heilprin, Tracy, Farquhar and Melville Glaciers and (b) Sharp, Hubbard, Bowdoin, Verhoeff, Morris Jesup and Diebitsch Glaciers. The numbers indicate the relative contributions of the summer frontal ablation (red) to the annual ablation in %. The data are the mean values over a period between September 2014 and September 2016. (c) Scatter plots of the summer frontal ablation and its relative contribution vs the annual frontal ablation rate.

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use. Journal of Glaciology 33

fjord circulation driven by subglacial meltwater discharge Author contribution. DS processed satellite data, analyzed the results and (e.g. Sciascia and others, 2013). We speculate that the contribu- wrote the paper. SS contributed to the interpretation of the results and writing. tion of summer submarine melting is more significant at these glaciers because warm AW is not available in winter. Submarine References melting increases in summer and enhances calving (Fried and others, 2015; Rignot and others, 2015), which results in greater Amundson JM and 5 others (2010) Ice mélange dynamics and implications seasonal variations in frontal ablation. for terminus stability, Jakobshavn Isbræ, Greenland. Journal of Geophysical Research 115, F01005. doi: 10.1029/2009JF001405. Andrews LC and 7 others (2014) Direct observations of evolving subglacial drainage beneath the Greenland ice sheet. Nature 514(7520), 80–83. doi: 10.1038/nature13796. 6. Conclusions Bartholomew I and 5 others (2010) Seasonal evolution of subglacial drainage – We studied ice-speed variations of 10 marine-terminating outlet and acceleration in a Greenland outlet glacier. Nature Geoscience 3(6), 408 glaciers along the coast of Prudhoe Land in northwestern 411. doi: 10.1038/ngeo863. Carr JR, Stokes CR and Vieli A (2013) Recent progress in understanding Greenland. During the study period from 2014 to 2016, the gla- marine-terminating Arctic outlet glacier response to climatic and oceanic ciers showed a common seasonal speed pattern: speedup began forcing: twenty years of rapid change. Progress in Physical Geography 37 between late May and early June, increased gradually until it (4), 436–467. doi: 10.1177/0309133313483163. reached its peak value between late June and early July, and sub- CReSIS (2016) Radar Depth Sounder L2 Data, Lawrence, Kansas, USA, Digital sequently slowed down from July to September. The magnitude of Media. Available at http://data.cresis.ku.edu/. − the speedup was within the range of 54 to 278 m a 1 (8–226%). Fountain AG and Walder JS (1998) Water flow through temperate glaciers. This seasonal speed pattern was similar to the one reported in Reviews of Geophysics 36(3), 299–328. doi: 10.1029/97RG03579. the nearby Melville Bay region in northwest Greenland (74.92– Fried MJ and 8 others (2015) Distributed subglacial discharge drives signifi- cant submarine melt at a Greenland tidewater glacier. Geophysical 76.08°N) (Moon and others, 2014), suggesting that the observed – seasonal speed pattern is commonly distributed over all of north- Research Letters 42, 9328 9336. doi: 10.1002/2015GL065806. Heid T and Kääb A (2012) Evaluation of existing image matching methods for western Greenland. deriving glacier surface displacements globally from optical satellite Observed summer speedups began approximately when hourly imagery. Remote Sensing of Environment 118, 339–355. doi: 10.1016/j.rse. air temperatures exceeded 0°C in early summer. Ice speed decel- 2011.11.024. erated in midsummer when air temperature still remained at the Howat IM, Box JE, Ahn Y, Herrington A and McFadden EM (2010) highest level. The influence of frontal retreat and sea-ice/ Seasonal variability in the dynamics of marine-terminating outlet glaciers ice-mélange conditions was insignificant. Based on these observa- in Greenland. Journal of Glaciology 56(198), 601–613. doi: 10.3189/ tions, we conclude that summer speedup was driven by subglacial 002214310793146232. water pressure which increased due to the meltwater input to the Howat IM, Joughin I, Fahnestock M, Smith BE and Scambos TA (2008) glacier bed, and the glacier decelerated when the pressure dropped Synchronous retreat and acceleration of southeast Greenland outlet glaciers through the development of an efficient subglacial drainage sys- 2000-06: ice dynamics and coupling to climate. Journal of Glaciology 54 (187), 646–660. doi: 10.3189/002214308786570908. tem. The contribution of summer speedup to annual ice motion Iken A (1972) Measurement of water pressure in moulins as part of a movement −1 – ranged from 2 to 62 m a (0.5 13%), and the relative contribu- study of the White Glacier, Axel Heiberg Island, Northwest Territories, tion was greater at slow-flowing glaciers. Canada. Journal of Glaciology 77,53–58. doi: 10.3189/S0022143000022486. Frontal ablation of the study glaciers showed seasonal varia- Iken A and Bindschadler RA (1986) Combined measurements of subglacial tions as well. Greater ablation was observed in the summer period water pressure and surface velocity of Findelengletscher, Switzerland: con- (June–August) and the seasonality is more enhanced in relatively clusions about drainage system and sliding mechanism. Journal of small and slow-flowing glaciers. For example, more than a half of Glaciology 32, 101–119. doi: 10.3189/S0022143000006936. the annual frontal ablation occurred during the summer period at Joughin I and 7 others (2008a) Continued evolution of Jakobshavn Isbrae Bowdoin, Verhoeff, Morris Jesup and Diebitsch Glaciers. These following its rapid speedup. Journal of Geophysical Research 113, F04006. glaciers showed clear seasonal variations in the front position: doi: 10.1029/2008JF001023. Joughin I and 5 others (2008b) Seasonal speedup along the western flank of advance from March to June/July and retreat from July to the Greenland Ice Sheet. Science 320(5877), 781–783. doi: 10.1126/sci- September. The frontal variations were dominated by the seasonal ence.1153288. variations in the frontal ablation rather than by ice-speed varia- Joughin I, Smith BE, Shean DE and Floricioiu D (2014) Brief communica- tions, suggesting the importance of ocean influence on calving tion: further summer speedup of Jakobshavn Isbræ. Cryosphere 8(1), 209– and submarine melting. Observed seasonal front variations were 214. doi: 10.5194/tc-8-209-2014. critically controlled by summer frontal ablation particularly at King MD and 6 others (2018) Seasonal to decadal variability in ice discharge relatively small slow-flowing glaciers. At the front of these glaciers, from the Greenland ice sheet. Cryosphere 12, 3813–3825. doi: 10.5194/ calving is enhanced by speedup and sea-ice/ice-mélange weaken- tc-12-3813-2018. ing during the summer period. Lemos A, Shepherd A, McMillan M and Hogg AE (2018) Seasonal variations in the flow of land-terminating glaciers in Central-West Greenland using Supplementary material. The supplementary material for this article can sentinel-1 imagery. Remote Sensing 10, 1878. doi: 10.3390/rs10121878. be found at https://doi.org/10.1017/jog.2019.81. Mankoff KD and 10 others (2019) Greenland ice sheet solid ice discharge from 1986 through 2017. Earth System Science Data 11, 769–786. doi: Acknowledgements. We thank Jason Amundson and an anonymous referee 10.5194/essd-11-769-2019. for constructive review comments. The paper was handled by the Scientific McNabb RW and Hock R (2014) Alaska tidewater glacier terminus positions, Editor Shad O’Neel and commented by the Chief Editor Hester Jiskoot. The 1948–2012. Journal of Geophysical Research. Earth Surface 119. doi: Landsat images were downloaded from the United States Geological Survey 10.1002/2013JF002915. Earth Explorer (http://earthexplorer.usgs.gov/). The dataset of air temperature Moon T and 6 others (2014) Distinct patterns of seasonal Greenland glacier was downloaded from the National Oceanic and Atmospheric velocity. Geophysical Research Letters 41. doi: 10.1002/2014GL061836. Administration’s National Centers for Environmental Information (https:// Moon T and Joughin I (2008) Changes in ice front position on Greenland’s www.ncdc.noaa.gov/). English text was corrected by Susan Braun-Clarke. outlet glaciers from 1992 to 2007. Journal of Geophysical Research 113, This study was funded by MEXT Japan (Japanese Ministry of Education, F02022. doi: 10.1029/2007JF000927. Culture, Sports, Science and Technology) through the Arctic Challenge for Moon T, Joughin I and Smith B (2015) Seasonal to multiyear variability of Sustainability (ArCS) Project. glacier surface velocity, terminus position, and sea ice/ice mélange in

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use. 34 Daiki Sakakibara and Shin Sugiyama

northwest Greenland. Journal of Geophysical Research. Earth Surface 120, northwestern Greenland. Journal of Glaciology 64(244), 300–310. doi: 10. 818–833. doi: 10.1002/2015JF003494. 1017/jog.2018.20. Moon T, Joughin I, Smith B and Howat I (2012) 21st-century evolution Scambos TA, Dutkiewicz MJ, Wilson JC and Bindschadler RA (1992) of Greenland outlet glacier velocities. Science 336(6081), 576–578. doi: Application of image cross-correlation to the measurement of glacier ice 10.1126/science.1219985. speed using satellite image data. Remote Sensing of Environment 42, 177– Nick FM, Vieli A, Howat IM and Joughin I (2009) Large-scale changes in 186. doi: 10.1016/0034-4257(92)90101-O. Greenland outlet glacier dynamics triggered at the terminus. Nature Schoof C (2010) Ice-sheet acceleration driven by melt supply variability. Geoscience 2(2), 110–114. doi: 10.1038/ngeo394. Nature 468(7325), 803–806. doi: 10.1038/nature09618. OMG Mission (2016) Bathymetry (sea floor depth) data from the ship-based Sciascia R, Straneo F, Cenedese C and Heimbach P (2013) Seasonal variability bathymetry survey. Ver. 0.1. OMG SDS, CA, USA. Available at https://doi. of submarine melt rate and circulation in an East Greenland fjord. Journal of org/10.5067/OMGEV-BTYSS (Accessed 2 February 2018). Geophysical Research. Oceans 118, 2492–2506. doi: 10.1002/jgrc.20142. Pimentel S and 7 others (2017) Modelling intra-annual dynamics of a major Sole AJ and 6 others (2011) Seasonal speedup of a Greenland marine- marine-terminating Arctic glacier. Annals of Glaciology 58(74), 118–130. terminating outlet glacier forced by surface melt-induced changes in subgla- doi: 10.1017/aog.2017.23. cial hydrology. Journal of Geophysical Research 116, F03014. doi: 10.1029/ Porter DF and 6 others (2014) Bathymetric control of tidewater glacier 2010JF001948. mass loss in northwest Greenland. Earth and Planetary Science Letters Sugiyama S and Gudmundsson GH (2003) Diurnal variations in vertical 401,40–46. doi: 10.1016/j.epsl.2014.05.058. strain observed in a temperate glacier. Geophysical Research Letters Rignot E and 9 others (2016) Bathymetry data reveal glaciers vulnerable to 30(2), 1090. doi: 10.1029/2002GL016160. ice-ocean interaction in Uummannaq and Vaigat glacial fjords, west Sugiyama S, Sakakibara D, Tsutaki S, Maruyama M and Sawagaki T (2015) Greenland. Geophysical Research Letters 43. doi: 10.1002/2016GL067832. Glacier dynamics near the calving front of Bowdoin Glacier, northwestern Rignot E, Fenty I, Xu Y, Cai C and Kemp C (2015) Undercutting of marine- Greenland. Journal of Glaciology 61(226), 223–232. doi: 10.3189/ terminating glaciers in West Greenland. Geophysical Research Letters 42, 2015JoG14J127. 5909–5917. doi: 10.1002/2015GL064236. Sundal AV and 5 others (2011) Melt-induced speed-up of Greenland ice sheet Rignot E and Kanagaratnam P (2006) Changes in the velocity structure of the offset by efficient subglacial drainage. Nature 469, 521–524. doi: 10.1038/ Greenland Ice Sheet. Science 311, 986–990. doi: 10.1126/science.1121381. nature09740. Ryser C and 7 others (2014) Caterpillar-like ice motion in the of Tsutaki S, Sugiyama S, Sakakibara D and Sawagaki T (2016) Surface elevation the Greenland ice sheet. Journal of Geophysical Research. Earth Surface, 119, changes during 2007–13 on Bowdoin and Tugto Glaciers, northwestern 2258–2271. doi: 10.1002/2013JF003067. Greenland. Journal of Glaciology 62(236), 1083–1092. doi: 10.1017/jog.2016.106. Sakakibara D and Sugiyama S (2014) Ice-front variations and speed changes van den Broeke MR and 8 others (2016) On the recent contribution of the of calving glaciers in the Southern Patagonia Icefield from 1984 to 2011. Greenland ice sheet to sea level change. Cryosphere 10(5), 1933–1946. Journal of Geophysical Research. Earth Surface 119, 2541–2554. doi: 10. doi: 10.5194/tc-10-1933-2016. 1002/2014JF003148. Vijay S and 5 others (2019) Resolving seasonal ice velocity of 45 Greenlandic Sakakibara D and Sugiyama S (2018) Ice front and flow speed variations of glaciers with very high temporal details. Geophysical Research Letters 46, marine-terminating outlet glaciers along the coast of Prudhoe Land, 1485–1495. doi: 10.1029/2018GL081503.

Downloaded from https://www.cambridge.org/core. 30 Sep 2021 at 14:27:26, subject to the Cambridge Core terms of use.